The AliveCor KardiaBand, a sensor that is compatible with the Apple Watch is able to detect dangerous levels of potassium in the blood with an accuracy of 94%. This is an interesting step forward seeing as, at the moment, the condition is usually caught with the use of blood tests which require needles.
The KardiaBand by AliveCor is a sensor that snaps into a slot on the watchband. The user touches the sensor; it is the sensor that takes a reading of the electrical activity of the heart called electrocardiogram (EKG). This reading is able to expose abnormal heart rhythm and atrial fibrillation (AFib), the sensor then sends the data to an app. At the American College of Cardiology conference in Florida yesterday, the CEO of AliveCor, Vic Gundotra presented research done with the Mayo Clinic showing that the same technology could detect too-high levels of potassium in the blood (hyperkalemia).
Hyperkalemia can be caused by diabetes, dehydration, and chronic kidney disease amongst other conditions. It could cause heart or kidney failure and it has no obvious symptoms – you can it one without knowing.
Excessive potassium interferes with the electrical activity of cells including the cells of the heart. According to Gundotra, this entails danger to the heart, but also, high potassium levels change the electrical reading of the heart – certain EKG patterns can reveal the presence of too much potassium. AliveCor worked alongside Mayo Clinic to develop a new algorithm for the KardiaBand which had the ability to analyze EKG data and determine if the user has hyperkalemia. The dataset had 2 million EKGs linked to 4 million potassium values collected over 23 years.
In order to train AI with these data points, the team divided the dataset into various parts. Some of the data was used to train the network to recognize the EKG patterns that revealed hyperkalemia and the AI had to learn for itself how to detect the pattern. When they completed the training, the team tested the AI on different parts of the data to check if, with the right EKG, they could tell if it showed hyperkalemia. The results displayed about 90 to 94% accuracy.
However, it will take time to see this technology since according to Gundotra, the FDA is yet to clear KardiaBand for diagnosing hyperkalemia.